Title: A Preference Programming Approach to Make the Even Swaps Method Even Easier
1A Preference Programming Approach to Make the
Even Swaps Method Even Easier
- Jyri Mustajoki
- Raimo P. Hämäläinen
- Systems Analysis Laboratory
- Helsinki University of Technology
- www.sal.hut.fi
2Outline
- The Even Swaps method
- Hammond, Keeney and Raiffa (1998, 1999)
- A new combined Even Swaps / Preference
Programming approach - PAIRS method (Salo and Hämäläinen, 1992)
- Additive MAVT model of the problem
- Intervals to model incomplete information
- Support for different phases of the Even Swaps
process - Smart-Swaps Web software
- The first software for supporting the method
3Even Swaps
- Multicriteria method to find the best alternative
- An even swap
- A value trade-off, where a consequence change in
one attribute is compensated with a comparable
change in some other attribute - A new alternative with these revised consequences
is equally preferred to the initial one - ? The new alternative can be used instead
4Elimination process
- Carry out even swaps that make
- Alternatives dominated (attribute-wise)
- There is another alternative, which is equal or
better than this in every attribute, and better
at least in one attribute - Attributes irrelevant
- Each alternative has the same value on this
attribute - ? These can be eliminated
- Process continues until one alternative, i.e.
the best one, remains
5Practical dominance
- If alternative y is slightly better than
alternative x in one attribute, but worse in all
or many other attributes - ? x practically dominates y
- ? y can be eliminated
- Aim to reduce the size of the problem in obvious
cases - Eliminate unnecessary even swap tasks
6Example
- Office selection problem (Hammond et al. 1999)
An even swap
7Supporting Even Swaps with Preference Programming
- Even Swaps process carried out as usual
- The DMs preferences simultaneously modeled with
Preference Programming - Intervals allow us to deal with incomplete
information about the DMs preferences - Trade-off information given in the even swaps can
be used to update the model - ? Suggestions for the Even Swaps process
- Generality of assumptions of Even Swaps preserved
8Supporting Even Swaps with Preference Programming
- Support for
- Identifying practical dominances
- Finding candidates for the next even swap
- Both tasks need comprehensive technical screening
- Idea supporting the process not automating it
9Decision support
10Assumptions in the Preference Programming model
- Additive value function
- Not a very restrictive assumption
- Weight ratios and component value functions are
initially within some reasonable bounds - General bounds for these often assumed
- E.g. practical dominance implicitly assumes
reasonable bounds for the weight ratios
11Preference Programming The PAIRS method
- Imprecise statements with intervals on
- Attribute weight ratios (e.g. 1/5 ? w1 / w2 ? 5)
- ? Feasible region for the weights
- Alternatives ratings (e.g. 0.6 ? v1(x1) ? 0.8)
- ? Intervals for the overall values
- Lower bound for the overall value of x
- Upper bound correspondingly
12Initial assumptions produce bounds
- For the weight ratios
- For the ratings
- Modeled with exponential value functions
- Any monotone value functions within the bounds
allowed - Additional bounds for the
min/max slope
13Use of trade-off information
- With each even swap the user reveals new
information about her preferences - This trade-off information can be utilized in the
process - ? Tighter bounds for the weight ratios obtained
from the given even swaps - ? Better estimates for the values of the
alternatives
14Practical dominance
- An alternative which is practically dominated
cannot be made non-dominated with any reasonable
even swaps - Analogous to pairwise dominance concept in
Preference Programming
15Pairwise dominance
- x dominates y in a pairwise sense if
- i.e. if the overall value of x is greater than
the one of y with any feasible weights of
attributes and ratings of alternatives - ? Any pairwisely dominated alternative can be
considered to be practically dominated
16Candidates for even swaps
- Aim to make as few swaps as possible
- Often there are several candidates for an even
swap - In an even swap, the ranking of the alternatives
may change in the compensating attribute - ? One cannot be sure that the other alternative
becomes dominated with a certain swap
17Applicability index
- Assume y is better than x only in attribute i
- Applicability index of an even swap, where a
change xi?yi is compensated in attribute j, to
make y dominated - Indicates how close to making y dominated we can
get with this swap - The bigger d is, the more likely it is to reach
dominance
18Applicability index
- Ratio between
- The minimum feasible rating change in the
compensating attribute to reach dominance and - The maximum possible rating change that could be
made in this attribute - Worst case value for d
- Bounds include all the possible impecision
- Average case value for d
- Rating differences from linear value functions
- Weight ratios as averages of their bounds
19Example
Initial Range 85 - 50 A - C 950 - 500 1500 -1900
36 different options to carry out an even swap
that may lead to dominance E.g. change in Monthly
Cost of Montana from 1900 to 1500 Compensation
in Client Access d(M?B, Cost, Access)
((85-78)/(85-50)) / ((1900-1500)/(1900-1500))
0.20 d(M?L, Cost, Access) ((85-80)/(85-50))
/ ((1900-1500)/(1900-1500)) 0.14 Compensation
in Office Size d(M?B, Cost, Size)
((950-500)/(950-500)) / ((1900-1500)/(1900-1500))
1.00 d(M?L, Cost, Size) ((950-700)/(950-500
)) / ((1900-1500)/(1900-1500)) 0.56
(Average case values for d used)
20Comparison with MAVT
Even Swaps MAVT
Assumptions about the value function Not needed Needed Additive functions typically used
Elicitation burden No. of elicitations may become high Not known in advance Increases with the no. of alternatives Weight elicitation At least n-1 preference statements Value functions One for each attribute
21Comparison with MAVT
Even Swaps MAVT
Analysis of the results Dominance relations No relative scores Outcomes of the alternatives change during the process Overall scores for the alternatives Clear to interpret
Suitability Personal decision making Proposed approach makes the process easier Group and policy decisions Transparency of the process
22Smart-Swaps softwarewww.smart-swaps.hut.fi
- Identification of practical dominances
- Suggestions for the next even swap to be made
- Additional support
- Information about what can be achieved with
each swap - Notification of dominances
- Rankings indicated by colors
- Process history allows backtracking
23Problem definition
24Entering trade-offs
25Process history
26www.Decisionarium.hut.fi
- Software for different types of problems
- Smart-Swaps (www.smart-swaps.hut.fi)
- Opinions-Online (www.opinions.hut.fi)
- Global participation, voting, surveys group
decisions - Web-HIPRE (www.hipre.hut.fi)
- Value tree based decision analysis and support
- Joint Gains (www.jointgains.hut.fi)
- Multi-party negotiation support
- RICH Decisions (www.rich.hut.fi)
- Rank inclusion in criteria hierarchies
27Conclusions
- Modeling of the DMs preferences in Even Swaps
with Preference Programming allows to - Identify practical dominances
- Find candidates for even swaps
- Makes the Even Swaps process even easier
- Support provided as suggestions by the
Smart-Swaps software
28References
- Hämäläinen, R.P., 2003. Decisionarium - Aiding
Decisions, Negotiating and Collecting Opinions on
the Web, Journal of Multi-Criteria Decision
Analysis, 12(2-3), 101-110. - Hammond, J.S., Keeney, R.L., Raiffa, H., 1998.
Even swaps A rational method for making
trade-offs, Harvard Business Review, 76(2),
137-149. - Hammond, J.S., Keeney, R.L., Raiffa, H., 1999.
Smart choices. A practical guide to making better
decisions, Harvard Business School Press, Boston. - Mustajoki, J., Hämäläinen, R.P., 2005. A
Preference Programming Approach to Make the Even
Swaps Method Even Easier, Decision Analysis,
2(2), 110-123. - Salo, A., Hämäläinen, R.P., 1992. Preference
assessment by imprecise ratio statements,
Operations Research, 40(6), 1053-1061. - Applications of Even Swaps
- Gregory, R., Wellman, K., 2001. Bringing
stakeholder values into environmental policy
choices a community-based estuary case study,
Ecological Economics, 39, 37-52. - Kajanus, M., Ahola, J., Kurttila, M., Pesonen,
M., 2001. Application of even swaps for strategy
selection in a rural enterprise, Management
Decision, 39(5), 394-402.